consul/agent/proxycfg/ingress_gateway.go

299 lines
10 KiB
Go
Raw Normal View History

// Copyright (c) HashiCorp, Inc.
// SPDX-License-Identifier: MPL-2.0
package proxycfg
import (
"context"
"fmt"
agent: remove agent cache dependency from service mesh leaf certificate management (#17075) * agent: remove agent cache dependency from service mesh leaf certificate management This extracts the leaf cert management from within the agent cache. This code was produced by the following process: 1. All tests in agent/cache, agent/cache-types, agent/auto-config, agent/consul/servercert were run at each stage. - The tests in agent matching .*Leaf were run at each stage. - The tests in agent/leafcert were run at each stage after they existed. 2. The former leaf cert Fetch implementation was extracted into a new package behind a "fake RPC" endpoint to make it look almost like all other cache type internals. 3. The old cache type was shimmed to use the fake RPC endpoint and generally cleaned up. 4. I selectively duplicated all of Get/Notify/NotifyCallback/Prepopulate from the agent/cache.Cache implementation over into the new package. This was renamed as leafcert.Manager. - Code that was irrelevant to the leaf cert type was deleted (inlining blocking=true, refresh=false) 5. Everything that used the leaf cert cache type (including proxycfg stuff) was shifted to use the leafcert.Manager instead. 6. agent/cache-types tests were moved and gently replumbed to execute as-is against a leafcert.Manager. 7. Inspired by some of the locking changes from derek's branch I split the fat lock into N+1 locks. 8. The waiter chan struct{} was eventually replaced with a singleflight.Group around cache updates, which was likely the biggest net structural change. 9. The awkward two layers or logic produced as a byproduct of marrying the agent cache management code with the leaf cert type code was slowly coalesced and flattened to remove confusion. 10. The .*Leaf tests from the agent package were copied and made to work directly against a leafcert.Manager to increase direct coverage. I have done a best effort attempt to port the previous leaf-cert cache type's tests over in spirit, as well as to take the e2e-ish tests in the agent package with Leaf in the test name and copy those into the agent/leafcert package to get more direct coverage, rather than coverage tangled up in the agent logic. There is no net-new test coverage, just coverage that was pushed around from elsewhere.
2023-06-13 15:54:45 +00:00
"github.com/hashicorp/consul/agent/leafcert"
"github.com/hashicorp/consul/agent/proxycfg/internal/watch"
"github.com/hashicorp/consul/agent/structs"
Protobuf Refactoring for Multi-Module Cleanliness (#16302) Protobuf Refactoring for Multi-Module Cleanliness This commit includes the following: Moves all packages that were within proto/ to proto/private Rewrites imports to account for the packages being moved Adds in buf.work.yaml to enable buf workspaces Names the proto-public buf module so that we can override the Go package imports within proto/buf.yaml Bumps the buf version dependency to 1.14.0 (I was trying out the version to see if it would get around an issue - it didn't but it also doesn't break things and it seemed best to keep up with the toolchain changes) Why: In the future we will need to consume other protobuf dependencies such as the Google HTTP annotations for openapi generation or grpc-gateway usage. There were some recent changes to have our own ratelimiting annotations. The two combined were not working when I was trying to use them together (attempting to rebase another branch) Buf workspaces should be the solution to the problem Buf workspaces means that each module will have generated Go code that embeds proto file names relative to the proto dir and not the top level repo root. This resulted in proto file name conflicts in the Go global protobuf type registry. The solution to that was to add in a private/ directory into the path within the proto/ directory. That then required rewriting all the imports. Is this safe? AFAICT yes The gRPC wire protocol doesn't seem to care about the proto file names (although the Go grpc code does tack on the proto file name as Metadata in the ServiceDesc) Other than imports, there were no changes to any generated code as a result of this.
2023-02-17 21:14:46 +00:00
"github.com/hashicorp/consul/proto/private/pbpeering"
)
type handlerIngressGateway struct {
handlerState
}
func (s *handlerIngressGateway) initialize(ctx context.Context) (ConfigSnapshot, error) {
snap := newConfigSnapshotFromServiceInstance(s.serviceInstance, s.stateConfig)
// Watch for root changes
err := s.dataSources.CARoots.Notify(ctx, &structs.DCSpecificRequest{
Datacenter: s.source.Datacenter,
QueryOptions: structs.QueryOptions{Token: s.token},
Source: *s.source,
}, rootsWatchID, s.ch)
if err != nil {
return snap, err
}
// Get information about the entire service mesh.
err = s.dataSources.ConfigEntry.Notify(ctx, &structs.ConfigEntryQuery{
Kind: structs.MeshConfig,
Name: structs.MeshConfigMesh,
Datacenter: s.source.Datacenter,
QueryOptions: structs.QueryOptions{Token: s.token},
EnterpriseMeta: *structs.DefaultEnterpriseMetaInPartition(s.proxyID.PartitionOrDefault()),
}, meshConfigEntryID, s.ch)
if err != nil {
return snap, err
}
// Watch this ingress gateway's config entry
err = s.dataSources.ConfigEntry.Notify(ctx, &structs.ConfigEntryQuery{
Kind: structs.IngressGateway,
Name: s.service,
Datacenter: s.source.Datacenter,
QueryOptions: structs.QueryOptions{Token: s.token},
EnterpriseMeta: s.proxyID.EnterpriseMeta,
}, gatewayConfigWatchID, s.ch)
if err != nil {
return snap, err
}
// Watch the ingress-gateway's list of upstreams
err = s.dataSources.GatewayServices.Notify(ctx, &structs.ServiceSpecificRequest{
Datacenter: s.source.Datacenter,
QueryOptions: structs.QueryOptions{Token: s.token},
ServiceName: s.service,
EnterpriseMeta: s.proxyID.EnterpriseMeta,
}, gatewayServicesWatchID, s.ch)
if err != nil {
return snap, err
}
snap.IngressGateway.WatchedDiscoveryChains = make(map[UpstreamID]context.CancelFunc)
snap.IngressGateway.DiscoveryChain = make(map[UpstreamID]*structs.CompiledDiscoveryChain)
snap.IngressGateway.WatchedUpstreams = make(map[UpstreamID]map[string]context.CancelFunc)
snap.IngressGateway.WatchedUpstreamEndpoints = make(map[UpstreamID]map[string]structs.CheckServiceNodes)
snap.IngressGateway.WatchedGateways = make(map[UpstreamID]map[string]context.CancelFunc)
snap.IngressGateway.WatchedGatewayEndpoints = make(map[UpstreamID]map[string]structs.CheckServiceNodes)
snap.IngressGateway.WatchedLocalGWEndpoints = watch.NewMap[string, structs.CheckServiceNodes]()
2021-07-13 12:53:59 +00:00
snap.IngressGateway.Listeners = make(map[IngressListenerKey]structs.IngressListener)
snap.IngressGateway.UpstreamPeerTrustBundles = watch.NewMap[string, *pbpeering.PeeringTrustBundle]()
snap.IngressGateway.PeerUpstreamEndpoints = watch.NewMap[UpstreamID, structs.CheckServiceNodes]()
snap.IngressGateway.PeerUpstreamEndpointsUseHostnames = make(map[UpstreamID]struct{})
return snap, nil
}
func (s *handlerIngressGateway) handleUpdate(ctx context.Context, u UpdateEvent, snap *ConfigSnapshot) error {
if u.Err != nil {
return fmt.Errorf("error filling agent cache: %v", u.Err)
}
switch {
case u.CorrelationID == rootsWatchID:
roots, ok := u.Result.(*structs.IndexedCARoots)
if !ok {
return fmt.Errorf("invalid type for response: %T", u.Result)
}
snap.Roots = roots
case u.CorrelationID == gatewayConfigWatchID:
resp, ok := u.Result.(*structs.ConfigEntryResponse)
if !ok {
return fmt.Errorf("invalid type for response: %T", u.Result)
}
if resp.Entry == nil {
return nil
}
gatewayConf, ok := resp.Entry.(*structs.IngressGatewayConfigEntry)
if !ok {
return fmt.Errorf("invalid type for config entry: %T", resp.Entry)
}
snap.IngressGateway.GatewayConfigLoaded = true
snap.IngressGateway.TLSConfig = gatewayConf.TLS
if gatewayConf.Defaults != nil {
snap.IngressGateway.Defaults = *gatewayConf.Defaults
}
2021-07-13 12:53:59 +00:00
// Load each listener's config from the config entry so we don't have to
// pass listener config through "upstreams" types as that grows.
for _, l := range gatewayConf.Listeners {
key := IngressListenerKeyFromListener(l)
2021-07-13 12:53:59 +00:00
snap.IngressGateway.Listeners[key] = l
}
if err := s.watchIngressLeafCert(ctx, snap); err != nil {
return err
}
case u.CorrelationID == gatewayServicesWatchID:
services, ok := u.Result.(*structs.IndexedGatewayServices)
if !ok {
return fmt.Errorf("invalid type for response: %T", u.Result)
}
// Update our upstreams and watches.
var hosts []string
watchedSvcs := make(map[UpstreamID]struct{})
upstreamsMap := make(map[IngressListenerKey]structs.Upstreams)
for _, service := range services.Services {
u := makeUpstream(service)
uid := NewUpstreamID(&u)
// TODO(peering): pipe destination_peer here
watchOpts := discoveryChainWatchOpts{
id: uid,
name: u.DestinationName,
namespace: u.DestinationNamespace,
2021-08-20 16:57:45 +00:00
partition: u.DestinationPartition,
datacenter: s.source.Datacenter,
}
up := &handlerUpstreams{handlerState: s.handlerState}
err := up.watchDiscoveryChain(ctx, snap, watchOpts)
if err != nil {
return fmt.Errorf("failed to watch discovery chain for %s: %v", uid, err)
}
watchedSvcs[uid] = struct{}{}
hosts = append(hosts, service.Hosts...)
id := IngressListenerKeyFromGWService(*service)
upstreamsMap[id] = append(upstreamsMap[id], u)
}
snap.IngressGateway.Upstreams = upstreamsMap
snap.IngressGateway.UpstreamsSet = watchedSvcs
snap.IngressGateway.Hosts = hosts
snap.IngressGateway.HostsSet = true
for uid, cancelFn := range snap.IngressGateway.WatchedDiscoveryChains {
if _, ok := watchedSvcs[uid]; !ok {
for targetID, cancelUpstreamFn := range snap.IngressGateway.WatchedUpstreams[uid] {
delete(snap.IngressGateway.WatchedUpstreams[uid], targetID)
delete(snap.IngressGateway.WatchedUpstreamEndpoints[uid], targetID)
cancelUpstreamFn()
targetUID := NewUpstreamIDFromTargetID(targetID)
if targetUID.Peer != "" {
snap.IngressGateway.PeerUpstreamEndpoints.CancelWatch(targetUID)
snap.IngressGateway.UpstreamPeerTrustBundles.CancelWatch(targetUID.Peer)
}
}
cancelFn()
delete(snap.IngressGateway.WatchedDiscoveryChains, uid)
}
}
if err := s.watchIngressLeafCert(ctx, snap); err != nil {
return err
}
default:
return (*handlerUpstreams)(s).handleUpdateUpstreams(ctx, u, snap)
}
return nil
}
// Note: Ingress gateways are always bound to ports and never unix sockets.
// This means LocalBindPort is the only possibility
func makeUpstream(g *structs.GatewayService) structs.Upstream {
upstream := structs.Upstream{
DestinationName: g.Service.Name,
DestinationNamespace: g.Service.NamespaceOrDefault(),
DestinationPartition: g.Service.PartitionOrDefault(),
LocalBindPort: g.Port,
IngressHosts: g.Hosts,
// Pass the protocol that was configured on the ingress listener in order
// to force that protocol on the Envoy listener.
Config: map[string]interface{}{
"protocol": g.Protocol,
},
}
return upstream
}
func (s *handlerIngressGateway) watchIngressLeafCert(ctx context.Context, snap *ConfigSnapshot) error {
// Note that we DON'T test for TLS.enabled because we need a leaf cert for the
// gateway even without TLS to use as a client cert.
if !snap.IngressGateway.GatewayConfigLoaded || !snap.IngressGateway.HostsSet {
return nil
}
// Watch the leaf cert
if snap.IngressGateway.LeafCertWatchCancel != nil {
snap.IngressGateway.LeafCertWatchCancel()
}
ctx, cancel := context.WithCancel(ctx)
agent: remove agent cache dependency from service mesh leaf certificate management (#17075) * agent: remove agent cache dependency from service mesh leaf certificate management This extracts the leaf cert management from within the agent cache. This code was produced by the following process: 1. All tests in agent/cache, agent/cache-types, agent/auto-config, agent/consul/servercert were run at each stage. - The tests in agent matching .*Leaf were run at each stage. - The tests in agent/leafcert were run at each stage after they existed. 2. The former leaf cert Fetch implementation was extracted into a new package behind a "fake RPC" endpoint to make it look almost like all other cache type internals. 3. The old cache type was shimmed to use the fake RPC endpoint and generally cleaned up. 4. I selectively duplicated all of Get/Notify/NotifyCallback/Prepopulate from the agent/cache.Cache implementation over into the new package. This was renamed as leafcert.Manager. - Code that was irrelevant to the leaf cert type was deleted (inlining blocking=true, refresh=false) 5. Everything that used the leaf cert cache type (including proxycfg stuff) was shifted to use the leafcert.Manager instead. 6. agent/cache-types tests were moved and gently replumbed to execute as-is against a leafcert.Manager. 7. Inspired by some of the locking changes from derek's branch I split the fat lock into N+1 locks. 8. The waiter chan struct{} was eventually replaced with a singleflight.Group around cache updates, which was likely the biggest net structural change. 9. The awkward two layers or logic produced as a byproduct of marrying the agent cache management code with the leaf cert type code was slowly coalesced and flattened to remove confusion. 10. The .*Leaf tests from the agent package were copied and made to work directly against a leafcert.Manager to increase direct coverage. I have done a best effort attempt to port the previous leaf-cert cache type's tests over in spirit, as well as to take the e2e-ish tests in the agent package with Leaf in the test name and copy those into the agent/leafcert package to get more direct coverage, rather than coverage tangled up in the agent logic. There is no net-new test coverage, just coverage that was pushed around from elsewhere.
2023-06-13 15:54:45 +00:00
err := s.dataSources.LeafCertificate.Notify(ctx, &leafcert.ConnectCALeafRequest{
Datacenter: s.source.Datacenter,
Token: s.token,
Service: s.service,
DNSSAN: s.generateIngressDNSSANs(snap),
EnterpriseMeta: s.proxyID.EnterpriseMeta,
}, leafWatchID, s.ch)
if err != nil {
cancel()
return err
}
snap.IngressGateway.LeafCertWatchCancel = cancel
return nil
}
// connectTLSServingEnabled returns true if Connect TLS is enabled at either
// gateway level or for at least one of the specific listeners.
func connectTLSServingEnabled(snap *ConfigSnapshot) bool {
if snap.IngressGateway.TLSConfig.Enabled {
return true
}
for _, l := range snap.IngressGateway.Listeners {
if l.TLS != nil && l.TLS.Enabled {
return true
}
}
return false
}
func (s *handlerIngressGateway) generateIngressDNSSANs(snap *ConfigSnapshot) []string {
// Update our leaf cert watch with wildcard entries for our DNS domains as
// well as any configured custom hostnames from the service. Note that in the
// case that only a subset of listeners are TLS-enabled, we still load DNS
// SANs for all upstreams. We could limit it to only those that are reachable
// from the enabled listeners but that adds a lot of complication and they are
// already wildcards anyway. It's simpler to have one certificate for the
// whole proxy that works for any possible upstream we might need than try to
// be more selective when we are already using wildcard DNS names!
if !connectTLSServingEnabled(snap) {
return nil
}
var dnsNames []string
namespaces := make(map[string]struct{})
for _, upstreams := range snap.IngressGateway.Upstreams {
for _, u := range upstreams {
namespaces[u.DestinationNamespace] = struct{}{}
}
}
// TODO(partitions): How should these be updated for partitions?
for ns := range namespaces {
// The default namespace is special cased in DNS resolution, so special
// case it here.
if ns == structs.IntentionDefaultNamespace {
ns = ""
} else {
ns = ns + "."
}
dnsNames = append(dnsNames, fmt.Sprintf("*.ingress.%s%s", ns, s.dnsConfig.Domain))
dnsNames = append(dnsNames, fmt.Sprintf("*.ingress.%s%s.%s", ns, s.source.Datacenter, s.dnsConfig.Domain))
if s.dnsConfig.AltDomain != "" {
dnsNames = append(dnsNames, fmt.Sprintf("*.ingress.%s%s", ns, s.dnsConfig.AltDomain))
dnsNames = append(dnsNames, fmt.Sprintf("*.ingress.%s%s.%s", ns, s.source.Datacenter, s.dnsConfig.AltDomain))
}
}
dnsNames = append(dnsNames, snap.IngressGateway.Hosts...)
return dnsNames
}